199 AI Tool Notes, Six Months, 211 Clicks: The Full Search Data

The short answer

Between 24 February and 23 August 2026 this site ran 199 notes about new AI tools, one most days, in English and Spanish. That's 398 URLs. 192 of them were published inside that window and 7 were already up when it started.

They earned 16,031 impressions and 211 clicks. A click-through rate of 1.32%, or about 1.2 clicks a day.

The interesting part isn't that the number is small. It's that every intermediate metric looked fine. Notes got indexed in a median of two days. They ranked positions 6 to 10 on hundreds of queries. By every proxy the strategy was working, and it still produced almost nothing, for a reason that took a while to see.

Here's all of it, including the parts that don't flatter anyone.

How this was measured

Every number comes from the Search Console API for sc-domain:ac0.ai, pulled on 23 August 2026 for the window 24 February to 23 August 2026, which is 181 days. Dimensions saved: date, query, page, query by page, page by date, country and device.

Publish dates come from the publishedAt field in each note's source file, so "how long until it ranked" is a join between the repo and Search Console rather than a guess.

Four things to hold in mind before trusting any of it:

• This is one site, one niche, one window. It's evidence, not a law.
• Positions in Search Console are averages, which flatten a lot of variance.
• The branded-versus-generic split later on uses a classifier I wrote: a query counts as branded if it contains a word that appears in one of the note slugs. It's approximate at the edges.
• Query-level totals are smaller than site totals on purpose, which turns out to be a finding in itself.

Everything on the site that isn't a field note, so the services pages, case studies, toolkit and home page combined, produced 764 impressions and 5 clicks in the same window. The notes are effectively the whole picture.

Finding 1: indexing was never the problem

The thing people worry about first turned out to be the thing that worked best.

Across 169 notes published inside the window, the median time from publishing to the first recorded impression was 2 days. 72% appeared within a week. 82% within a month.

Percentile	Days to first impression
p10	1
p25	1
p50	2
p75	10
p90	60

A Next.js site on Vercel with a clean sitemap and a daily publishing rhythm gets picked up quickly. If you're currently worrying about crawl budget or submission tooling on a small site, this is weak evidence that you're worrying about the wrong thing.

Finding 2: almost all of the value arrives in week one

Grouping every daily impression by how old that note was on that day:

Note age	Impressions	Clicks
0 to 7 days	6,349	133
8 to 30 days	3,356	34
31 to 90 days	4,893	31
91+ days	1,720	8

About 65% of all clicks arrived in a note's first week. Past 90 days, 1,720 impressions produced 8 clicks.

That's the opposite of how evergreen content is supposed to behave, and it makes sense once you see what the traffic actually is. A note about a tool released on Tuesday catches the burst of people looking that tool up on Tuesday, Wednesday and Thursday. Then the burst ends, because interest in a specific new repo doesn't compound.

Finding 3: the click curve is broken in a specific place

This is the number that reframed everything else.

Position bucket	Queries	Impressions	Clicks	CTR
1 to 3	20	176	22	12.50%
4 to 5	26	175	0	0.00%
6 to 10	188	3,572	52	1.46%
11 to 20	48	149	2	1.34%
21 to 50	84	300	0	0.00%
51+	52	170	0	0.00%

Positions 4 and 5 produced 175 impressions and not a single click.

For comparison, published CTR benchmarks put position 4 around 7.2% and position 10 around 1.6%, in a meta-analysis dated 11 August 2026. Worth noting that these benchmarks disagree with each other quite badly, since another 2026 roundup puts position 1 at 27% where the first puts it at 39.8%. Treat them as a rough shape rather than a target.

Even against the generous reading, this site's position 6 to 10 bucket underperforms the published position 10 rate. Ranking on page one was worth roughly nothing.

Finding 4: why, and it's not a fixable content problem

Search the name of a tool the site covers and look at what's there.

For "t3mp3st", the results are led by the GitHub repo itself, then a YouTube tutorial, then a handful of aggregator posts describing the same repo. The note here sits at an average position of 6.5. It's the site's single best page: 1,110 impressions, 50 clicks.

For "webwright", the results are led by Microsoft Research, the GitHub repo and the official docs site. The note here averages position 9.0, and turned 1,233 impressions into 7 clicks.

That's the mechanism. Somebody typing a tool's name is performing a navigational search. They want that tool's own page, and it's sitting directly above you, and it's free, and it's authoritative. Being the sixth-best answer to "where is this repo" is a position with no economic value at any level of writing quality.

It also explains the position 4 and 5 result. Those are ranks just below an official page, on queries where the official page is the answer. The reason 74% of note slugs got zero clicks isn't that the notes are bad. It's that they were entered into a race whose prize goes to someone else.

Finding 5: two thirds of the impressions come from queries Google won't name

The site total is 16,031 impressions. Adding up every named query gives 5,153.

The gap is 10,878 impressions, 67.9% of the total, from queries Google refuses to show. This is documented behaviour: Search Console omits anonymized queries, meaning ones not issued by more than a few dozen users over a two to three month period. They count in the totals and never appear in the table.

Publishing about brand-new tools generates an unusual amount of this. The searches are things like a repo name plus a typo, or a half-remembered project plus "github". Real people, individually so rare that Google won't put them in writing.

Among the queries it will name, the split is stark:

	Queries	Impressions	Clicks	CTR	Weighted position
Branded, tool-name lookups	386	4,895	81	1.65%	10.5
Generic	65	258	1	0.39%	21.4

Six months of daily publishing produced 258 impressions of non-branded demand. One click.

Finding 6: the audience is real, and it isn't the buyer

Splitting by country:

Country	Impressions	Clicks	CTR	Avg position
United States	3,823	16	0.42%	14.3
Spain	1,658	35	2.11%	9.4
India	1,161	23	1.98%	9.0
Mexico	567	8	1.41%	8.3
Chile	479	9	1.88%	7.1
United Kingdom	477	5	1.05%	8.7
Indonesia	372	23	6.18%	7.7

Indonesia converts at fifteen times the US rate on a tenth of the impressions. The Spanish-language half of the site, which is cheap to produce since every note is written in both languages, does 4,674 impressions and 69 clicks against English's 11,644 and 137, at a better CTR.

This is a genuinely engaged developer audience across Spain, Latin America and South and Southeast Asia. It's just nobody who is going to hire a European B2B consultancy, which is what the business here actually sells. The content found an audience. It found the wrong one, and no amount of editing fixes a geography and intent mismatch.

What the concentration looks like

Slice	Share of note impressions
Top 1 URL	8.7%
Top 5 URLs	33.4%
Top 10 URLs	45.5%
Top 20 URLs	58.9%
Top 50 URLs	79.2%

46% of slugs finished under 20 lifetime impressions. 74% got no clicks at all. 19% of note URLs got at least one click.

The median note earned 23 impressions in six months. The mean is dragged up entirely by a handful of pages that happened to cover a tool that got popular.

What I'd tell someone doing the same thing

Three things, and none of them are "write better".

Check what the winning result is before you write. If a query's best answer is a GitHub repo, an official docs page or a vendor's own site, you're competing for a click that has already been spent. Coverage of a specific named product is nearly always this. It's a real cost and it's invisible until you look at the SERP rather than the keyword.

Watch the age curve, not the total. A rising impression count on a daily publishing schedule mostly measures how recently you published. Splitting by note age was the single most useful thing in this analysis, because a growing total was hiding the fact that individual notes were dying in a week.

Separate the queries you can name from the ones you can't. If most of your impressions are anonymized, you don't have a keyword strategy, you have a lottery with good odds of small prizes. That's a fine thing to run deliberately. It's a bad thing to run by accident while believing you're building topical authority.

The two deliberate comparison pieces the site has published, on headless browsers for AI agents and AI tools that can see your screen, were written against generic queries instead of product names. The first is 20 days old and sits at an average position of 32.0 with 261 impressions. Too early to call, and the fair thing is to say so and check again at 90 days rather than claim a result now.

What the six months does say clearly is that the archive has value, just not the value that was assumed. 199 dated notes on tools most people hadn't heard of yet is a record of having been early, and being early is checkable. That's worth something in a conversation. It was never going to be worth much in a search result.

Words worth knowing

Impression. Your page appeared in someone's search results. They didn't necessarily see it or scroll to it.

CTR, click-through rate. Clicks divided by impressions. 1.32% means about one search in seventy-six ended in a visit.

Navigational query. A search where the person already knows where they want to go, like typing a product's name to find its site. The destination almost always wins the click.

Anonymized query. A search too rare for Google to name in Search Console, because too few people made it. It counts in your totals but you'll never see the words.

Average position. Where you typically ranked, averaged across every impression. A page at "position 8" was often higher and lower than that.

Content decay. A page earning less over time. On this site it was fast, with most value gone after the first week.